ChatGPT or human teachers? Exploring feedback comprehension, use, and perceptions in L2 writing classrooms

The growing demand for English skills has led to large classes and teachers being unable to give personalized feedback to everyone. Large language models such as ChatGPT offer a potential solution to this issue by providing students with more access to feedback. However, there is limited data showing the ability of this technology to adequately meet the needs of second language (L2) learners seeking to develop their writing skills. To address this gap, this study compared the feedback from ChatGPT with that of a human teacher in terms of perceived usefulness and effect on the academic writing of low-intermediate L2 university students. Within a quasi-experimental design, Group A received ChatGPT feedback, and Group B received teacher feedback. Each group completed three writing tasks over six weeks. Data were gathered using blind scoring of drafts, success on revision, and semi-structured interviews. The results showed that students exposed to teacher feedback had greater comprehension, revised more successfully, and believed more in the feedback compared to students who received feedback from ChatGPT. Although ChatGPT appeared useful in fixing surface-level concerns, such as spelling and grammar, large language models could complement traditional teacher feedback, especially where resources are limited.